Software failures that do not repeatedly occur have been solved by acquiring and analyzing the processing log of the software to identify the cause of the failures.
Processing log acquisition according to such prior-art methods has the following problems:
(1) Software modules must be modified to add a routine for acquiring a process log. Accordingly, the workload for acquiring the process log is heavy.
(2) Processing log acquisition is performed for each module, and therefore logs are generated on a module-by-module basis. It is difficult to obtain a chronological log of the processing of the entire software (the software as a whole). Consequently, a large number of man-hours are required for analyzing a log to identify the cause of a failure because the log does not provide a broad view of the entire processing.